A check-in is usually a set of sliders. Mood, energy, sleep quality, pain — each rated out of ten, daily. It produces a tidy chart and surprisingly little that anyone can act on.
The reason is that a number records how something seemed without recording what happened, and what happened is what a clinician needs.
The four things worth recording
1. What actually happened, in your own words. One sentence. Not a rating.
Could not start anything until after lunch. Once I did, it was fine. Lost the thread twice in the team meeting, which people noticed. Good day. Read a chapter and remembered it.
This is the most valuable field by a distance, and it is the one most systems leave out. A sentence carries the kind of information — which capacity failed, in what context, with what consequence — that a number cannot represent.
2. The one or two things you are actually following. Not twelve. If the question is about afternoon clarity, record afternoon clarity. Everything else is noise you will have to read past.
A rating is fine here, as long as the question stays identical every time. Changing the wording makes the earlier entries incomparable.
3. Whether you took your medication, and when. Two seconds, and it prevents the single most common misinterpretation there is. A poor week with four missed doses and a poor week with none are completely different findings, and without this field they look identical.
4. The obvious confounders. Sleep, roughly. Whether the day was unusual. Alcohol, if relevant. Cycle day, if relevant.
Four items. That is the whole thing, and it is a minute.
Why free text beats a slider
It records the kind of failure, not just the amount. "Lost the thread mid-sentence" and "could not start anything" are both bad days and they point in different directions. As a 4 out of 10 they are identical.
It records consequence. Whether it cost you anything is one of the main determinants of urgency, and there is no slider for it.
It preserves context. "Bad day, but I had three hours of sleep and a migraine" is a different fact from "bad day for no reason". Without the clause, both are a 3.
It is readable later. Six weeks on, "Tuesday: 4/10" tells you nothing. A sentence tells you what happened.
The slider is not useless — a consistent number over weeks makes a trend visible. It works as a supplement to a sentence, not as a replacement for one.
What not to collect
More than a handful of fields. Long forms get abandoned, and abandoned unevenly — people complete them on bad days — which biases everything.
Things you cannot answer consistently. "Sleep quality" means something different every time you are asked. Time to bed and rough hours are answerable.
Anything you will not use. Every field has a cost in attention, and an unused field costs without returning anything.
A self-assessment you are worried about. If rating your own memory makes you anxious, record instances instead — what happened, not how bad it is.
Timing
Once a day, at the same time. Evening usually works, because the day has happened. Morning check-ins record the night rather than the day.
Or not daily at all. For most questions, event-based is better: write something when something happens. This avoids the noise of recording twenty ordinary days to catch four notable ones.
More around a change. The two or three weeks after a treatment change are the one clear case for daily.
What a month of this looks like
3rd — took it 8am. Fine morning, started work straight away. Flat from about 3. Slept 7h. 4th — took it 8am. Same. Slept 7h. 5th — missed it. Could not get going all day. Slept 6h. 6th — took it 10am, late. Afternoon worse than usual, headache from 4. Slept 5h, woke early. 9th — good day, best in a while. Read properly for the first time in weeks. Slept 8h.
Twelve lines like that answer questions nothing else can: the effect arrives, it wears off mid-afternoon, late dosing produces a headache, the good day followed the long night, and the worst day was a missed dose rather than a loss of effect.
None of that is visible in thirty daily numbers.
The review is where the value is
A log nobody reads is a diary. Before an appointment, spend five minutes turning it into a paragraph:
Three weeks. Consistent except four missed doses. Effect reliable in the morning, gone by around 3, and there is a flat patch after that. Headaches on the days I took it late. The best days followed the longest nights. Sleep has averaged about six hours.
That is what gets acted on. Bring the log as backup, but hand over the paragraph.
The honest boundary
What a check-in record establishes is what happened and what happened alongside it. Not why.
That the good days followed long nights is an association in your own data — a good reason to look at sleep, never proof that sleep was the cause. Keeping that boundary is what makes the record trustworthy, including to you.